FELIX: FEderated Test-beds for Large-scale Infrastructure
eXperiments
Bartosz Belter
bartosz.belter@man.poznan.pl
Focused Technical Workshop: International OpenFlow/SDN Testbeds
Florida International University
31 March – 2 April 2015
FELIX in a nutshell
Facts
Objectives
• EC (EU) & NICT (JP) collaborative project
- Collaboration between European and
Japanese partners
• Started in April 2013. To run till March 2016
• To createa large-scale testbed (experimental
infrastructure), federated across two
continents
• To define a common software architecture for
testbeds
EU coordinator
JP coordinator
NL
BE
PL
DE
JP
IT
ES
Federating existing SDN-based testbeds in EU and JP
Existing testbeds
• OFELIA (OpenFlow in Europe: Linking Infrastructure
and Applications)
First large-scale OpenFlow-enabled experimental
facility in Europe), continuously evolving through other
FIRE projects (FIBRE, Fed4FIRE, ALIEN, etc.)
www.fp7-ofelia.eu
• RISE (Research Infrastructure for large-Scale network
Experiments)
OpenFlow testbed over JGN-X (the largest testbed
network in Japan), with wide-area coverage from US
West coast to Southeast Asia
www.jgn.nict.go.jp/rise/english
Why to federate?
• To increase mutual benefits of European and Japanese researchers
by creating more complex environments for specialized research and
experiments
• This opens up new opportunities for experiments due to
geographical dispersion of testbeds
FELIX Use Cases – Towards a FELIX Architecture
Two major clusters of use cases
Based on different applicability area and stakeholders, use cases are divided into:
• Data Domain use cases
Optimize the use of the interconnectivity between testbeds to realize data transfer
1. Data on Demand – delivery of distributed data by setting data flows
2. Pre-processing and delivery of nearly real-time [satellite] data
3. High quality media transmission over long-distance networks
• Infrastructure Domain use cases
Optimize the use of the infrastructure as a whole, including the migration of entire data
processing workloads
4. Data mobility service by SDN technologies
5. Follow the Sun/Moon – green energy in data centers
6. Disaster recovery by migrating IaaS to a remote data center
FELIX Data Domain – Use Case #1
Data on Demand – delivery of distributed data by setting data flows over the network
A user wants to run an algorithm at a given site (Data Processor) over different data sets contained in
data storage locations around the world.
Research question. How can we efficiently retrieve large amounts of data stored along different
and distributed sites and process it on demand?
• Key entity: Data Processor (algorithms
could be proprietary) to control the
testbed network elements and define
the flow paths
• Key solution aspect: Dynamic
infrastructure interconnection and
reconfiguration through network
resource managers to set up NSI and
OpenFlow paths
FELIX Data Domain – Use Case #2
Pre-processing and delivery of nearly real-time [satellite] data to distant locations
Satellite sources generate huge amounts of nearly real-time data, later to be transmitted to a number
of distant locations interconnected by transit networks.
Research question. Can we reduce the size
of data to be delivered across the transit
network to improve the overall performance?
• Key entity: FELIX middleware to allocate
the proper computing and caching
resources at data source and destination(s)
ends, also to configure the network
segments on-demand
• Key solution aspect: Coordinated control
of computing and network resources
through resource orchestration and the
involved resource managers
FELIX Data Domain – Use Case #3
High Quality Media Transmission over long-distance networks
The federation of SDN testbeds interconnected through NSI-enabled domains opens up opportunities
for testing media streaming over existing network technologies in backbone networks (DWDM,
MPLS, etc.) in conjunction with an OpenFlow control plane deployed in the testbeds.
Research question. Can we determine the behavior of the transmission mechanisms when streaming
high resolution media content at very long distances?
• Key entities: Experiment control and performance monitoring
• Key solution aspect: NSI, OpenFlow control and monitoring for automatic adjustment of
connection path(s) through path parameters, depending on media streams and QoS & QoE
Experimenter
Domain A
Media Consumers
(End users)
Domain B
Testbed operator
Domain C
Media provider
FELIX Infrastructure Domain – Use Case #1
Data Mobility Service by SDN Technologies (Inter-Cloud use case)
A user of a service provided by a cloud system moves to a remote location (e.g. due to a business
trip).
Research question. Can the cloud system monitor performance and “move” selected data closer to
the remote location?
• Key entity: User experience
monitoring and manager
• Key solution aspects:
Monitoring (detection of the
“closest” location that
guarantees required service
quality, availability &
performance), NSI and
OpenFlow managers (configure
paths and flows), resource
management (migration of user
data, services and applications)
FELIX Infrastructure Domain – Use Case #2
Follow the Sun/Moon
Some considerations on energy consumption:
• Internet usage curves following a similar daily pattern around the world
• Cooling is reported to amount up to 50% of the total energy bill of data centers
• Prices of renewable energy may strongly depend on the availability of wind and solar energy
Research question. How can we move
the compute workflow to the nearest &
“greenest” data center available in a
federation?
• Key entity: Workload control and
management
• Key solution aspects: Monitoring
(detect energy cost for operation),
OpenFlow & NSI managers for
dynamic, on-demand, end-to-end
network connections and resource
management for (re-)provisioning
FELIX Infrastructure Domain – Use Case #3
Disaster recovery by migrating IaaS to a remote data center
After some disaster occurs, it will be highly beneficial to immediately continue the service at a
remote site,even if the data is slightly obsolete – up to hours or days. BCP: Business continuity is
sometimes more important than data integrity.
Research question. Can the entire IaaS platform (cluster of
VMs & management software) be migrated to remote,
differently configured data centers for business continuity?
• Key entity: Hardware as a Service (HaaS) coordination
software to configure and provide virtual resources by using
technologies such as KVM/XEN, OpenFlow, etc.
• Key solution aspects: NSI managers to configure the transit
links, resource management for migration
FELIX Use Cases – Recap
Why use cases?
• To use as a guide for the architecture definition
- Translate into requirements
- Help to identify architectural components
• To validate the features of the testbed
More details?
See paper & deliverable D2.1
www.ict-felix.eu
What shall FELIX provide to networking experimenters?
• Access to a large SDN experimental facility between locations from Europe and Japan
• Applicability of similar use cases
What shall FELIX architecture provide to testbed facility administrators?
• Federation with other GENI-enabled testbeds
• Easier deployment of new internal GENI-based resource managers
• Improved access control and resource monitoring
The FELIX Architecture (1)
•
•
•
The slice concept is adopted in FELIX
– Experimental facilities to be provided dynamically on top of the FELIX physical
infrastructure (federated testbeds).
All the experimental facilities will be controlled programmatically
– Facilities are composed of computing and network resources (CR and NR)
belonging to distributed SDN islands in FELIX infrastructure
– Orchestrate resources in a multi-domain environment
– In a slice, facilities are interconnected via TNS-controlled domains (transit
network).
User has access and control of a slice provided.
The FELIX Architecture (2)
Request configuration
of slice(s)
•
The FELIX Space provides
users with slices for their own
use. Users request slices to an
RO.
– RO: Resource Orchestrator
– RM: Resource Manager
– PHY RES: physical
resources (testbed)
•
The User Space consists of any
tools and applications that a user
wants to deploy to control a slice
or execute particular operations
Users
Manage slice
Control slice
Resource controllers in User Space may
communicate with FELIX Space
components
The FELIX Architecture (3)
KEY FEATURES
Resource Orchestration
•Orchestration of various virtualized resources (compute, network and storage) provided by multiple domains
Domain Resource Management
•Coordination of various resources provided by heterogeneous resource management systems within a domain
Resource Allocation Planning
•Consider both computing and network resources, user and resource administrator aspects, such as, for instance, cost,
energy consumption and load balancing
Provisioning
•A virtual flat environment, just like a dedicated cluster, using dynamic resource information
Authentication and Authorization
•All actions are performed (only) by authenticated actors
•Ensure all actors are who they claim to be; keep track of all authenticated action
Monitoring
•Information from multiple domains is aggregated and provided to the user
User Access/GUI
•Allow for easy interaction between the experimenter and the FELIX testbed federation
Inter-domain networking through NSI
RO: Resource Orchestrator, MRO: Master RO, VM: Virtual Machine, AM: Aggregate Manager,
RM: Resource Manager, SE-RM: Stitching Entity-RM, TN-RM: Transit Network-RM
Relevant Standards for Interoperability and Portability
OpenGridForum (OGF) http://www.ogf.org
– Network Services Interface Workign Group (NSI-WG)
•
•
•
An interface to request a multi-domain dynamic network service
Being currently standardized in OGF
To be deployed by R&E networks in Europe, Asia and US
In FELIX: A means to achieve inter-domain connectivity for SDN Islands, contribution to SDN
extensions in NSI
– Network Modeling Language Working Group (NML-WG)
•
a standardised network description ontology and schema, facilitating interoperability between
different research projects and existing initatives
In FELIX: ontology for SDN and NSI resources
Internet Research Task Force (IRTF) https://irtf.org/
– Software-Defined Networking Research Group (SDNRG)
•
•
•
•
Classification of SDN models (Definitions, Taxonomies, Relationship to work ongoing in the IETF
and other SDOs)
SDN model scalability and applicability
Multi-layer programability and feedback control systems
Network description languages, abstractions, interfaces and compilers
In FELIX: contribution to multi-domain SDN architectures, interfaces and description lang.
Achievements to date & Future plans
Architecure
• Six project use cases have been identified and
described (September 2013)
• FELIX architecture has been released
(February 2014)
Implementation
• Prototype implementations of the architectural
framework released (January 2015)
– Software page:
http://www.ict-felix.eu/wp-content/uploads/2014/03/FELIX-D2.1.pdf
http://www.ict-felix.eu/?page_id=824
– GitHub:
https://github.com/dana-i2cat/felix/wiki
•
FELIX testbed deployed
Experimental validation
• FELIX experiments and implementation of
project use cases just started (Q1-2015)
http://www.ict-felix.eu/wpcontent/uploads/2014/03/FELIX_D2.2_General_Architecture_and_Functional_Blocks.pdf
PARTNERS
Poznan Supercomputing
and Networking Center
Poland
National Institute
of Advanced Industrial Science
and Technology
Japan
Nextworks
Italy
Fundacio Privada i2CAT,
Internet I Innovacio Digital
A Catalunya
Spain
SURFnet bv
Netherlands
European Center for Information and
Communication Technologies Gmbh
Germany
iMinds VZW
Belgium
KDDI
Japan